DDevelopment & Design

Hibernate annotations + Spring transactions in OSGI

  • bympr
  • May 23, 2010
  • 3 minute read

On our new project we decided to develop backend as services deployed on Servicemix 4.2. To make it easier for developers more familiar with ‘traditional’ web applications based on spring+hibernate stack it should (at least partially) resemble such application. Therefore, we want to have following bundles:

  • model –  containing hibernate classes
  • dao – implemented using hibernate template
  • service – interfaces for bundles implementing webservices, also controlling transactions, validation and similar stuff

So far, so good. Additional requirements are:

  • hibernate is configured using annotations
  • we use declarative transaction management with spring @Transactional annotation

Now, the question is – how to achieve it in OSGI environment? There are some tutorials on the web, but I haven’t found anything about this particular configuration, so I want to share our solution – as it wasn’t as straightforward as one might expect.

Hibernate bundle

The very first problem was the structure of hibernate jars themselves. It turned out that simple

wrap:mvn:org.hibernate:hibernate:3.3.1.GA

is not enough when one wants to use annotations. The problem is that both hibernate-core, and hibernate-annotations are using same package, namely org.hibernate.cfg. This is pretty anti-OSGI design, as OSGI bundle can use package exported by exactly one bundle.

The solution is of course pretty simple – we have to create our own bundle, which embeds both hibernate jars. This can be achieved with following maven-bundle-plugin configuration:

          

    *;uses:="org.hibernate";version=3.3.1.GA
    !*
    *;scope=compile|runtime;type=!pom;inline=true

Of course, you have to add hibernate dependencies to pom.

Session factory and transaction management

Servicemix has its own JTA TransactionManager, and at first I wanted to use it for managing our hibernate transactions. Unfortunatelly it turned to be somewhat tricky, so I decided to use simple solution, i.e. org.springframework.orm.hibernate.HibernateTransactionManager. It is declared in our dao bundle and exported as OSGI service:

	

    org.springframework.transaction.PlatformTransactionManager

Then, in service bundle we import it, and configure declarative transactions:



So far, it doesn’t look much more complicated than in traditional, monolithic war application. But there is one more caveat:


org.hibernate.jdbc.ConnectionWrapper is not visible from class loader


At first I thought it’s a matter of some misconfigured import/export declarations in some of our bundles. However, after some googling it turned out that it’s Hibernate’s bug


It is caused by using Thread context classloader when obtaining connection. It looks that the bug itself is already fixed, but we didn’t want to change to Hibernate 3.5 just because of that.


It turns out that there is quire simple workaround – you just need to make Spring aspect that will set context classloader to proper one (i.e. bundle class loader) before opening hibernate session, and you’re done. But again, using Servicemix makes it slightly harded, as it turns out that using @AspectJ style poses some difficulties (see e.g. this thread).


But it turns out that there is pretty simple solution – use old, Spring 1.x-style aspects :). This requires a bit more boilerplate xml, but at least we got it working pretty quickly.


The aspect class, setting proper classloader:


public class AspectFix implements MethodInterceptor {

    public Object invoke(MethodInvocation invocation) throws Throwable {
        ClassLoader contextClassLoader = Thread.currentThread().getContextClassLoader();
        Thread.currentThread().setContextClassLoader(getClass().getClassLoader());
        try {
            return invocation.proceed();
        }
            finally{ Thread.currentThread().setContextClassLoader(contextClassLoader);
        }
    }


The xml config:



      .*

    *Service

      fix


Final remark – now the service bundle has to import org.hibernate.jdbc to make it all work – as its classloader is used to initialize connection now.


To sum up: this is definitely not the most elegant solution one can imagine- but it’s working.


		

		





	


			

			

				

					
				

				

					

						
							
						
					

					

									

			

		

	
	








	


				

			
	



					

						
					

						

							 — Previous article
						


						

							Introducing camel-drools component						

					

				
							
					

						

							Next article — 
						


						

							Apache ODE 1.3.4 released						

					

				
				

		


			


				
				

					

						You May Also Like					

				


				


					




	

		
DDevelopment & Design

		
		

						

				
Log4j and MDC in Grails
			


							

											

							
Log4j provides very useful feature: MDC - mapped diagnostic context. It can be used to store data in context of current thread. It may sound scary a bit but idea is simple.

My post is based on post http://burtbeckwith.com/blog/?p=521 from Burt Beckwith's excellent blog, it's definitely worth checking if you are interested in Grails.

Short background story...

Suppose we want to do logging our brand new shopping system and we want to have in each log customer's shopping basket number. And our system can be used at once by many users who can perform many transactions, actions like adding items and so on. How can we achieve that? Of course we can add basket number in every place where we do some logging but this task would be boring and error-prone. 

Instead of this we can use MDC to store variable with basket number in map. 

In fact MDC can be treated as map of custom values for current thread that can be used by logger. 


How to do that with Grails?

Using MDC with Grails is quite simple. All we need to do is to create our own custom filter which works for given urls and puts our data in MDC.

Filters in Grails are classes in directory grails-app/conf/* which names end with *Filters.groovy postfix. We can create this class manually or use Grails command: 
grails create-filters info.rnowak.App.Basket
 
In result class named BasketFilters will be created in grails-app/conf/info/rnowak/UberApp.

Initially filter class looks a little bit empty:
class BasketFilters {
    def filters = {
        all(controller:'*', action:'*') {
            before = {

            }
            after = { Map model ->

            }
            afterView = { Exception e ->

            }
        }
    }
}
All we need to do is fill empty closures, modify filter properties and put some data into MDC.

all is the general name of our filter, as class BasketFilters (plural!) can contain many various filters. You can name it whatever you want, for this post let assume it will be named basketFilter. 

Another thing is change of filter parameters. According to official documentation (link) we can customize our filter in many ways. You can specify controller to be filtered, its actions, filtered urls and so on. In our example you can stay with default option where filter is applied to every action of every controller. If you are interested in filtering only some urls, use uri parameter with expression describing desired urls to be filtered.

Three closures that are already defined in template have their function and they are started in these conditions:

  • before - as name says, it is executed before filtered action takes place
  • after - similarly, it is called after the action
  • afterView - called after rendering of the actions view
Ok, so now we know what are these mysterious methods and when they are called. But what can be done within them? In official Grails docs (link again) under section 7.6.3 there is a list of properties that are available to use in filter.

With that knowledge, we can proceed to implementing filter.

Putting something into MDC in filter

What we want to do is quite easy: we want to retrieve basket number from parameters and put it into MDC in our filter:
class BasketFilters {
    def filters = {
        basketFilter(controller:'*', action:'*') {
            before = {
                MDC.put("basketNumber", params.basketNumber ?: "")
            }
            after = { Map model ->
                MDC.remove("basketNumber")
            }
        }
    }
}
 
We retrieve basket number from Grails params map and then we put in map under specified key ("basketNumber" in this case), which will be later used in logger conversion pattern. It is important to remove custom value after processing of action to avoid leaks.

So we are putting something into MDC. But how make use of it in logs?

We can refer to custom data in MDC in conversion patter using syntax: %X{key}, where key is our key we used in filter to put data, like:
def conversionPattern = "%d{yyyy-MM-dd HH:mm:ss} %-5p %t [%c{1}] %X{basketNumber} - %m%n"
 

And that's it :) We've put custom data in log4j MDC and successfully used it in logs to display interesting values. 
						

					
									

			
		


	






	

		
DDevelopment & Design

		
		

						

				
Simple trick to DRY your Grails controller
			


							

											

							
Grails controllers are not very DRY. It's easy to find duplicated code fragments in default generated controller. Take a look at code sample below. It is duplicated four times in show, edit, update and delete actions:
 
class BookController {
def show() {
def bookInstance = Book.get(params.id)
if (!bookInstance) {
flash.message = message(code: 'default.not.found.message', args: [message(code: 'book.label', default: 'Book'), params.id])
redirect(action: "list")
return
        }
        [bookInstance: bookInstance]
    }
}
 
Why is it duplicated?
 
There is a reason for that duplication, though. If you move this snippet to a method, it can redirect to "list" action, but it can't prevent controller from further execution. After you call redirect, response status changes to 302, but after method exits, controller still runs subsequent code.
 
Solution
 
At TouK we've implemented a simple trick to resolve that situation:
 
  1. wrap everything with a simple withStoppingOnRender method,
  2. whenever you want to render or redirect AND stop controller execution - throw EndRenderingException.
We call it Big Return - return from a method and return from a controller at once. Here is how it works:
 
class BookController {
def show(Long id) {
withStoppingOnRender {
Book bookInstance = Book.get(id)
validateInstanceExists(bookInstance)
            [bookInstance: bookInstance]
        }
    }

protected Object withStoppingOnRender(Closure closure) {
try {
return closure.call()
        } catch (EndRenderingException e) {}
    }

private void validateInstanceExists(Book instance) {
if (!instance) {
flash.message = message(code: 'default.not.found.message', args: [message(code: 'book.label', default: 'Book'), params.id])
redirect(action: "list")
throw new EndRenderingException()
        }
    }
}

class EndRenderingException extends RuntimeException {}
 
Example usage
 
For simple CRUD controllers, you can use this solution and create some BaseController class for your controllers. We use withStoppingOnRender in every controller so code doesn't look like a spaghetti, we follow DRY principle and code is self-documented. Win-win-win! Here is a more complex example:
 
class DealerController {
@Transactional
def update() {
withStoppingOnRender {
Dealer dealerInstance = Dealer.get(params.id)
validateInstanceExists(dealerInstance)
validateAccountInExternalService(dealerInstance)
checkIfInstanceWasConcurrentlyModified(dealerInstance, params.version)
            dealerInstance.properties = params
saveUpdatedInstance(dealerInstance)
redirectToAfterUpdate(dealerInstance)
        }
    }
}